https://issues.apache.org/bugzilla/show_bug.cgi?id=56119
--- Comment #22 from Sebb <[email protected]> --- Just discovered something else. I tried testing timeouts against the ASF server. Even if the idle timeout is exceeded, the samples don't fail. Debug shows that this is because the ASF server returns the following header: Keep-Alive: timeout=5, max=100 This is detected by the HC 4.x code: 2014/02/19 19:37:47:619 GMT [DEBUG] headers - << Keep-Alive: timeout=5, max=100 2014/02/19 19:37:47:619 GMT [DEBUG] headers - << Connection: Keep-Alive 2014/02/19 19:37:47:619 GMT [DEBUG] headers - << Content-Type: text/html; charset=utf-8 2014/02/19 19:37:47:697 GMT [DEBUG] BasicClientConnectionManager - Connection can be kept alive for 5000 MILLISECONDS 2014/02/19 19:39:17:712 GMT [DEBUG] BasicClientConnectionManager - Get connection for route {s}->https://www.apache.org 2014/02/19 19:39:17:712 GMT [DEBUG] BasicClientConnectionManager - Connection [id:0][route:{s}->https://www.apache.org][state:null] expired @ Wed Feb 19 19:37:52 GMT 2014 2014/02/19 19:39:17:712 GMT [DEBUG] DefaultClientConnection - Connection 0.0.0.0:4368<->192.87.106.229:443 closed 2014/02/19 19:39:17:712 GMT [DEBUG] DefaultClientConnectionOperator - Connecting to www.apache.org:443 Note that the connection is automatically dropped by the connection manager. I have just reviewed the logs from the private test plan, and these don't return a Keep-Alive header, so the HC code assumes the connection is not timed out. The log shows the following: BasicClientConnectionManager - Connection can be kept alive indefinitely This means that the HC code does not know when to drop the connection. I think it is probably a configuration error on the server; it really ought to tell the client what the keep-alive period is. It probably explains why this issue has not been reported by others as well. It gives a possible solution for the test errors - fix the server config! -- You are receiving this mail because: You are the assignee for the bug.
